Characterizing the 3-D structure of each building in the conterminous United States

Abstract

The three-dimensional building structure is crucial for understanding human activities and environmental interactions. Previous studies have estimated building height on a grid-scale, lacking height estimation results concurrently over large areas and at fine resolution. Here we mapped the height of each building in the conterminous US in 2020, using integrated multi-source datasets (i.e., satellite observations and geometry features). Our proposed model for height estimation is proven reliable (i.e., The model's R2 is 0.82, and RMSE is 3.35 m in the conterminous US) and superior in cross-validation with other existing datasets at different scales. We found an explicit pattern of buildings decaying from the urban core to rural areas, which is more noticeable in big cities (e.g., New York). Furthermore, we proposed the seamless and explicit Urban Canopy Parameters (UCPs) datasets, including six UCPs (e.g., building area fraction and height -to-width ratio) based on our height dataset. As key parameters used in climate models, our UCPs improved the fine-grained description of 3-D urban structures compared with those traditionally used. Overall, our building-by-building height dataset and UCP datasets reveal the significant heterogeneity of urban underlying surface, which can help socioeconomic and climatological urban studies.